Introduction

On MR imaging, gadolinium contrast leads to high signal intensity in areas of inflammation, thereby increasing the visibility of non-enhancing regions such as abscesses. Several studies have demonstrated that the inflammatory changes on MRI related to musculoskeletal infection, with or without coexisting abscesses, are more conspicuous with the use of intravenous gadolinium [1–5]. ACR guidelines and recent review articles continue to support the use of intravenous gadolinium in the work-up of musculoskeletal infections [6–9]. To our knowledge there has been little research to determine the effect of gadolinium on diagnostic accuracy in the MRI evaluation of musculoskeletal infection and abscess.

Gadolinium is considered a very safe contrast agent; however, allergic reactions, although rare, do occur [10]. Routine gadolinium administration in the setting of musculoskeletal infection should be re-examined, particularly in the context of the availability of high field strength magnets, newer coils and improved pulse sequences that can provide superior imaging resolution.

The purpose of this study is to compare the diagnostic utility of noncontrast and contrast MRI examinations in the work-up of osteomyelitis and soft-tissue musculoskeletal infections in children.

Materials and methods

Patient population

This retrospective study was approved by our institutional review board with waiver of informed consent. At our institution, a total of 147 gadolinium contrast-enhanced MRI examinations were performed in children from July 2003 through August 2008 for suspected musculoskeletal infection. Among the 147 MRI examinations, 50 were excluded because they were follow-up exams in children who had already been treated or undergone surgical intervention prior to the MRI exam. An additional 7 children were excluded because they did not have clinical or surgical follow-up for reference diagnosis. The 90 remaining contrast-enhanced MRI examinations (mean patient age, 9 years; range, 1 month–19 years) constituted the study group of pre- and post-gadolinium MRI examinations. There were 56 boys and 34 girls. Their MRI exams consisted of 70 (78%) lower extremity (20 femur, 11 knee, 11 tibia, 7 ankle, 21 foot), 12 (13%) pelvis, and 8 (9%) upper extremity (1 hand, 1 finger, 3 forearm, 1 shoulder, 2 elbow) exams.

Clinical data analysis

Final clinical diagnosis, surgical intervention, and hospitalization were abstracted from the medical records. Final diagnoses were categorized as osteomyelitis, septic arthritis, myositis and/or cellulitis, or noninfectious causes, on the basis of the discharge diagnosis and clinical follow-up. The average time for follow-up was 7.3 weeks. A child was considered to have a surgical intervention if a treatment procedure such as arthrotomy or incision and drainage of bone or soft tissue was performed. Diagnostic procedures such as joint, marrow, and soft-tissue aspiration performed after MRI were categorized as separate and distinct from surgical interventions.

In the data analysis for final clinical diagnosis, hierarchal diagnosis based on severity of infection was used with the following system: osteomyelitis > septic arthritis > myositis/cellulitis. For example, if a child had osteomyelitis and myositis as the final clinical diagnoses, the clinical diagnosis was considered to be osteomyelitis for data analysis.

Imaging technique

All MRI examinations were performed on a 1.5-T Intera (Philips Healthcare, Andover, MA), 1.5-T Advantage (GE Healthcare, Milwaukee, WI), or 3.0-T Achieva (Philips Healthcare). A dedicated torso coil was used for imaging the pelvis or long appendicular bones. A dedicated knee coil was used for imaging the knee. Elsewhere, flexible surface coils were used. The pre-contrast sequences for all children included biplane or triplane fluid-sensitive (T2-weighted with fat saturation, proton density-weighted with fat-saturation, or STIR), and single-plane T1-weighted non-fat-saturated sequences. A minority of children also had a single-plane susceptibility (either 2-D or 3-D gradient-recalled echo with fat suppression) sequence performed. Post-contrast sequences for all children were single-plane or biplane gadolinium-enhanced T1-weighted fat-saturated sequences (gadopentetate dimeglumine, Magnevist, Bayer Healthcare Pharmaceuticals, Montville, NJ, USA, 0.2 ml/kg, max 20 ml).

Image analysis pre- and post-contrast sequences

Consensus review of the MRI studies was performed by two radiologists, both with additional CAQ in pediatric radiology, one with additional pediatric musculoskeletal subspecialty training, with a combined 26 years of post-graduate experience, who were blinded to the final diagnosis. Pre- and post-contrast studies were reviewed for diagnostic criteria of osteomyelitis, septic arthritis, myositis/cellulitis, presence or absence of abscess, other diagnosis, or normal. Subset evaluation in those who underwent pelvic MRI was also performed. Diagnostic criteria are detailed below. Pre-contrast image consensus review was performed without review of post-contrast images; post-contrast consensus review included review of all pre- and post-contrast sequences.

The diagnosis of osteomyelitis was made if there was evidence of marrow infiltrative fluid intensity signal, juxta-cortical soft-tissue infiltrative fluid intensity signal, cloaca, sequestrum, or sinus tract. For pre-contrast imaging, abscess was defined as a focal hyperintense loculated fluid collection within the intramedullary or subperiosteal space on fluid-sensitive sequences. For post-contrast imaging, abscess was defined as a focal hyperintense loculated fluid intensity collection on fluid-sensitive sequences with absent central enhancement on post-contrast imaging.

The diagnosis of septic arthritis was made when there was qualitatively abnormally increased fluid within a joint with associated juxta-articular soft-tissue or marrow infiltrative fluid intensity signal along both sides of the joint. For post-contrast imaging, synovial enhancement had to be present. When osteomyelitis was accompanied by secondary septic arthritis (i.e. asymmetric edema with additional features of osteomyelitis affecting only one juxta-articular bone), the study was categorized as osteomyelitis. For pre-contrast imaging, abscess was defined as a hyperintense loculated fluid intensity collection on fluid-sensitive sequences within the juxta-articular soft tissues. For post-contrast imaging, abscess was defined as any area located within the juxta-articular soft tissues that was hyperintense on fluid-sensitive sequences with rim enhancement and absent central enhancement on post-gadolinium images.

The diagnosis of cellulitis/myositis was made when there was infiltrative fluid intensity signal within the subcutaneous fat or muscles in the absence of underlying septic arthritis or osteomyelitis. For pre-contrast imaging, abscess was defined as hyperintense loculated fluid intensity collection on fluid-sensitive sequences within subcutaneous fat or muscles and absent fat signal on T1-W sequences. For post-contrast imaging, abscess was defined as any loculated hyperintense region on fluid-sensitive sequences with absent fat signal intensity on T1-W sequences, which additionally demonstrated rim enhancement and absent central enhancement on post-gadolinium images.

In data analysis for final radiologic diagnosis based on consensus review, hierarchal diagnosis based on severity of infection was used with the following system: osteomyelitis > septic arthritis > myositis/cellulitis. For example, if a child had osteomyelitis and significant juxtacortical myositis on MRI, the final radiologic diagnosis was osteomyelitis. When an abscess was present, the abscess was measured in its largest 2-dimensional plane.

Statistical analysis

This is a retrospective descriptive study. Proportions of patients with the study endpoints are calculated. Proportions between groups are compared using the exact chi-square test or Fisher exact test as appropriate based upon the number of events for the endpoints. Proportions of paired binary data collected on the same subject, such as abscess (yes or no) diagnosed with MRI before and after the administration of gadolinium contrast agent are compared using the McNemar test. Mean and range are reported for continuous endpoints such as age and size of abscesses. Continuous endpoint such as size of abscesses is compared between groups using the Wilcoxon rank sum test. P values less than 0.05 are considered statistically significant. All tests are two-tailed. Statistical analyses are performed using the statistical package SAS for Windows (Version 9, Cary, NC, USA) and statistical analysis package R.

Results

Overview

Demographics and final clinical diagnoses of 90 children based on chart review are listed in Table 1. There were 35 children with osteomyelitis (38.9%), 33 with myositis/cellulitis (36.7%), and 6 with septic arthritis (6.6%). Sixteen children (17.8%) had a final non-infectious clinical diagnosis; nine of these children had no etiology for symptoms based on MRI or clinical follow-up, and seven had other final diagnoses including scalded skin syndrome, toxic synovitis (2), post-traumatic soft-tissue edema without infection (2), neonatal onset multisystem inflammatory disease, and post-traumatic osteoarthritis.

Table 1 Final reference diagnosis and demographics
Full size table

There were no patients with normal pre-contrast MRI studies who ultimately were diagnosed with musculoskeletal infection. Only two patients (2.2%) had a change in MRI diagnosis with the administration of gadolinium (1 cellulitis/myositis to osteomyelitis, 1 cellulitis/myositis to septic arthritis). Overall, in the imaging of musculoskeletal infection, the diagnostic performance of musculoskeletal MRI with and without intravenous gadolinium showed no statistical difference in the diagnosis of osteomyelitis, septic arthritis, and cellulitis/myositis (Table 2), with the exception of the detection of complicating abscesses and improving conspicuity of inflammation, detailed below.

Table 2 Diagnostic sensitivities and specificities of MRI with and without intravenous gadolinium
Full size table

Osteomyelitis

Among the 35 children with a final clinical diagnosis of osteomyelitis, addition of gadolinium made a difference in only 1 child (2.8%); 31 were correctly diagnosed by pre-contrast imaging alone; 1 child was correctly diagnosed as osteomyelitis after review of the gadolinium-enhanced images, while 3 diagnoses of osteomyelitis were missed on both pre- and post-gadolinium images, when compared to the final diagnoses suggested by chart review. There was no significant change in sensitivity and specificity of the MRI diagnosis of osteomyelitis by the addition of gadolinium-enhanced sequences (Table 2).

The child correctly diagnosed only after review of the contrast images was initially categorized as cellulitis and was correctly changed to osteomyelitis after review of gadolinium-enhanced images. Three false-negative diagnoses of osteomyelitis, as defined by chart review, were made based on both pre- and post-contrast imaging. One child was interpreted during consensus review as an avulsion injury. However, with clinical information available, this child had been officially interpreted as osteomyelitis at the time the study was performed; he subsequently underwent surgical debridement and was treated for presumed osteomyelitis with negative cultures. A second child was interpreted as septic arthritis based on pre- and post-contrast images. However, the official interpretation at the time of the study was osteomyelitis and the child was treated for presumed osteomyelitis, again with negative cultures. The third child was interpreted at consensus review as myositis without osteomyelitis. Again, the official interpretation at the time of the study was osteomyelitis and the child was treated for presumed osteomyelitis with negative cultures.

No false-positive diagnoses of osteomyelitis were made with pre-contrast imaging alone; however, two false-positive diagnoses of osteomyelitis were made after review of post-contrast imaging. The discharge diagnosis in one of these children was septic arthritis, and the discharge diagnosis in the second child was toxic synovitis.

For the evaluation of osteomyelitis complicated by abscess, pre-contrast sequences, compared to post-contrast images and final discharge diagnosis, correctly identified 11/23 abscesses (48%), missed 12 abscesses correctly diagnosed after review of post-contrast images, and overcalled 2 abscesses not diagnosed after review of post-contrast images. Seven abscesses identified on pre-contrast imaging were surgically treated, 0 were aspirated, and 4 were treated conservatively. Of the 12 abscesses identified only after gadolinium, 6 were surgically treated (Fig. 1), 1 was aspirated, and 5 were treated conservatively.

Fig. 1
figure 1

Osteomyelitis of the ischiopubic synchondrosis, surgically treated. a T2-W fat-saturated axial image demonstrates ischiopubic osteitis and juxta-cortical soft-tissue edema initially interpreted as osteomyelitis without co-existing abscess. However, on retrospective review, a subperiosteal abscess was present (arrow) and was better delineated on (b) a post-gadolinium T1-W fat-saturated axial sequence (arrow). Abscess was confirmed surgically

Full size image

Of the two children with the false-positive diagnoses of abscess made on pre-contrast images only, one had surgery with histologic evidence of chronic osteomyelitis and without evidence of abscess (Fig. 2), and the second was treated conservatively and did well on antibiotics alone.

Fig. 2
figure 2

Osteomyelitis of the pubic bone, surgically treated. a T2-W fat-saturated axial image demonstrates pubic bone osteitis and suggests small juxta-cortical abscesses within the obturator internis muscle (arrow). b Post-gadolinium T1-W fat-saturated axial sequence demonstrates diffuse muscle enhancement without underlying abscess. No abscess was identified operatively

Full size image

Septic arthritis

Among the six children with a final clinical diagnosis of septic arthritis, addition of gadolinium made a difference in only one child (16.6%); three were correctly diagnosed by pre-contrast imaging alone; one was correctly diagnosed as septic arthritis after review of the gadolinium-enhanced images (Fig. 3), while two diagnoses of septic arthritis were missed on both pre- and post-gadolinium images, when compared to the final diagnoses suggested by chart review. There was no significant change in sensitivity and specificity of the MRI diagnosis of septic arthritis by the addition of gadolinium-enhanced sequences (Table 2).

Fig. 3
figure 3

Pyogenic sacroiliitis initially diagnosed as myositis on pre-contrast sequences and correctly identified with the addition of post-contrast sequences. a T2-W fat-saturated coronal image demonstrates iliacus myositis (arrow). b STIR coronal image demonstrates a normal appearance to the SI joints. c Post-gadolinium T1-W fat-saturated axial image demonstrates subtle asymmetric enhancement of the left SI joint (arrowhead) and iliacus myositis (arrow), consistent with sacroiliitis

Full size image

The child correctly diagnosed only after review of the contrasted images was initially categorized as myositis and correctly changed to septic arthritis after review of gadolinium-enhanced images (Fig. 3). The two false-negative diagnoses of septic arthritis, as defined by chart review, were made based on pre- and post-contrast imaging. One child was diagnosed as myositis based on imaging review. The official radiology interpretation at the time of patient encounter was septic arthritis, and this was confirmed operatively and clinically. The second child was diagnosed as osteomyelitis based on imaging review. The official radiology interpretation at the time of patient encounter was osteomyelitis and septic arthritis. Based on operative findings and clinical follow-up, the child was thought to have septic arthritis.

There were two false-positive diagnoses of septic arthritis based on pre- and post-contrast imaging. In one of these children, noted in the osteomyelitis section, the official interpretation at the time the study had been performed was osteomyelitis and the child was treated for presumed osteomyelitis with negative cultures. In the second child the official interpretation at the time of patient encounter had also been septic arthritis but the child was clinically thought to have no orthopedic or infectious-related symptoms and was treated with ibuprofen alone and did well.

In the evaluation of septic arthritis complicated by juxta-articular abscess, one child was diagnosed with abscess, and this was identified only on post-contrast images. This child underwent joint aspiration only but did not have the abscess debrided.

Of the six children with septic arthritis, two underwent joint debridement, three underwent joint aspiration only, and one was treated conservatively with antibiotics.

Cellulitis/myositis

Among the 33 children with a final clinical diagnosis of cellulitis/myositis, all were correctly diagnosed by pre-contrast imaging alone and all were correctly diagnosed as cellulitis/myositis after review of the gadolinium-enhanced images. There was no significant change in sensitivity and specificity of the MRI diagnosis of cellulitis/myositis by the addition of gadolinium-enhanced sequences (Table 2).

There were seven false-positive diagnoses of cellulitis/myositis based on evaluation of pre- and post-contrast imaging. Two of these studies were officially interpreted as normal at the time of clinical encounter, and the children were discharged without a diagnosis of infection (one child was admitted with pneumonia and developed hip pain during hospital course and this resolved, the second child’s symptoms were thought to be related to trauma and not infection). One of the seven false-positive studies had been officially interpreted as cellulitis at the time of clinical encounter, but with further work-up the child was discharged with diagnosis of neonatal-onset multisystem inflammatory disease. One study had been officially interpreted as cellulitis, but clinically the edema was thought to be related to trauma and not infection. One study had been interpreted as cellulitis at the time of clinical encounter, but the child was discharged with the clinical diagnosis of scalded skin syndrome. One study was interpreted as myositis at the time of clinical encounter, and clinically was diagnosed as septic arthritis (as aforementioned). The last of these seven false-positive studies was interpreted as myositis and possibly early osteomyelitis at the time of clinical encounter, and clinically was diagnosed as osteomyelitis.

There were two false-positive diagnoses of cellulitis/myositis based on pre-contrast imaging alone, which were correctly diagnosed on post-contrast imaging. With the administration of gadolinium, these two children initially thought to have cellulitis/myositis were categorized as having more severe infections, as aforementioned (1 septic arthritis, 1 osteomyelitis).

In the evaluation of cellulitis/myositis complicated by abscess, pre-contrast imaging alone correctly identified 9/18 (50%) abscesses and missed 9 abscesses that were correctly diagnosed after review of post-contrast images in comparison to final discharge diagnoses (Fig. 4). Seven abscesses (78%) identified on pre-contrast imaging were surgically treated and two were treated conservatively; none was aspirated. Of the 9 abscesses identified only after gadolinium, 1 (11%) was treated surgically and 8 were treated conservatively; none was aspirated. The one abscess identified after gadolinium enhancement that required surgical drainage measured 1.1 cm2 in maximal axial dimension and was located in the prepatellar bursa.

Fig. 4
figure 4

Cellulitis of the knee, surgically treated. a T2-W fat-saturated axial image demonstrates diffuse cellulitis without co-existing abscess. b Post-gadolinium T1-W fat-saturated axial image at the same level demonstrates soft-tissue abscess (arrows) not initially appreciated on non-contrast sequences. The abscess was confirmed operatively

Full size image

Identification of musculoskeletal abscess

Based on pre-contrast imaging, 22 abscesses were prospectively identified and 15 (68.2%) of those required surgical intervention (Table 3). An additional 22 abscesses were identified based on post-contrast sequences, and of these 7 (31.8%) required surgical intervention; of these, 6 were abscesses related to osteomyelitis, and 1 was an abscess related to cellulitis/myositis. Two children with abscesses identified on pre-contrast imaging were thought to not have abscesses based on post-contrast imaging.

Table 3 Intra-osseous and extra-articular soft-tissue abscesses identified by MRI
Full size table

The average size of the 22 abscesses identified on pre-contrast images was 3.8 cm (min 0.1 cm, max 34.2 cm). The average size of the 20 additional abscesses found only on post-contrast images was 2.3 cm2 (min 0.2 cm2, max 12.6 cm2) (P = 0.53) (Fig. 5). The average size of abscesses that were treated non-surgically was 1.6 cm2 and the average of those that required surgical drainage was 5.0 cm2 (P = 0.005) (Table 4).

Fig. 5
figure 5

Extensive cellulitis of the foot with large abscess appreciated only after gadolinium administration. a T2-W fat-saturated long-axis image demonstrates diffuse infiltrative subcutaneous edema without loculated fluid collection to suggest abscess. b Post-gadolinium T1-W fat-saturated long-axis image clearly delineates rim-enhancing fluid collection confined to subcutaneous fat. Despite its large size, this abscess was treated non-operatively

Full size image
Table 4 Size of abscesses identified by MRI
Full size table

Subset analysis of pelvic MRI exams

A subset of children with suspected musculoskeletal infection who underwent pelvic MRI (n = 12) was evaluated. The final diagnoses for this subset of children were osteomyelitis 4, septic arthritis 3, cellulitis/myositis 0, and normal 5. There was one change in diagnosis based on pre- and post-contrast imaging. Based on pre-contrast imaging the diagnosis in one child was cellulitis/myositis and this changed to septic arthritis based on post-contrast imaging. Pre- and post-contrast pelvic MRI review independently and correctly identified all five children who were normal based on final diagnosis. Two abscesses related to osteomyelitis were identified based on pre-contrast imaging. Four abscesses were identified on post-contrast imaging, three related to osteomyelitis and one related to septic arthritis located within the juxta-articular soft tissues. Of the four abscesses, only one abscess, which was related to osteomyelitis, required surgical intervention, and this was only identified based on post-contrast imaging. All three cases of septic arthritis underwent joint aspiration.

The sensitivity and specificity of pre- and post-contrast diagnosis of osteomyelitis was 100%. The sensitivity and specificity of pre-contrast diagnosis of septic arthritis was 67% and 100%, respectively. The sensitivity and specificity of post-contrast diagnosis of septic arthritis was 100%.

Discussion

MRI is a complementary and clinically accepted test in the work-up of musculoskeletal infection because it can (1) confirm the diagnosis of infection, (2) provide a surgical roadmap to localize abscesses requiring debridement or to obtain a culture specimen, and (3) offer alternative etiologies for child’s symptoms when infection is not present [11, 12]. Based on our study, we have found that the routine use of gadolinium does not significantly alter our ability to diagnose musculoskeletal infection in children. Gadolinium-supplemented images did identify additional abscesses not appreciated on pre-contrast imaging and this was statistically significant. However, only a minority of these smaller abscesses required surgical debridement and they were most often related to underlying osteomyelitis.

The routine use of gadolinium in the work-up of musculoskeletal infection is problematic for several reasons. Administration of gadolinium increases costs and magnet time and can have potentially adverse affects in an already sick child. In children, studies must be performed quickly to decrease sedation time or avoid the need for sedation. For these reasons, we believe that the administration of gadolinium should not be routine on all children, particularly those with entirely normal pre-contrast examinations, but be decided thoughtfully, based on evidence of its effect on the accuracy of image interpretation.

We found that gadolinium administration is helpful when a study is abnormal to further define abscesses (Fig. 6). In some circumstances, gadolinium administration made musculoskeletal infections more conspicuous (Fig. 6). Despite the benefit and conspicuity of delineating infections, only two children (2.2%) had a change in diagnosis. In these two cases, more severe infections were identified after the administration of contrast agent (cellulitis/myositis upgraded to osteomyelitis and cellulitis/myositis upgraded to septic arthritis). No cases of musculoskeletal infections were missed and only identified after review of gadolinium-enhanced sequences. No children who had a final diagnosis of musculoskeletal infection had a normal exam based on pre-contrast imaging alone.

Fig. 6
figure 6

Osteomyelitis of the fourth metatarsal with abscess treated conservatively. a T2-W fat-saturated short-axis image demonstrates fourth metatarsal osteitis (arrow) and large plantar abscess (arrowhead). b Post-gadolinium T1-W fat-saturated short-axis image demonstrates fourth metatarsal osteomyelitis (arrow), more conspicuously compared with pre-contrast sequences. The plantar abscess (arrowhead) is equally conspicuous on both pre- and post-contrast sequences

Full size image

Additional abscesses identified with post-gadolinium sequences and requiring surgical intervention were most frequently seen in children with osteomyelitis. Of the 12 additional abscesses identified with post-gadolinium sequences in children with osteomyelitis, six (50%) required surgical intervention. Although additional abscesses were identified with post-gadolinium sequences in children with cellulitis/myositis, these rarely required surgical intervention. Of the nine additional abscesses identified with post-gadolinium sequences in patients with cellulitis/myositis, only one (11.1%) required surgical intervention.

Based on our findings, we believe that intravenous gadolinium does not need to be routinely administered in children who are referred for pelvic and extra-pelvic musculoskeletal MRI with suspected infection when pre-contrast imaging is entirely normal. However, in studies that are even mildly abnormal on pre-contrast imaging (Fig. 3), gadolinium-supplemented images might be helpful to rule out more severe infections and better define abscesses when present [12] or to facilitate properly identifying malignancy that can superficially mimic infection, such as Ewing sarcoma [13]. Therefore, to optimize patient care, contain cost and improve utilization, it is recommended that the radiologist screen studies prior to intravenous gadolinium administration to determine whether post-contrast imaging is necessary.

We found two studies that advocate for the limited use of gadolinium in the imaging work-up of musculoskeletal infection. Miller et al. [14], in a relatively small number of adult patients, showed that gadolinium does not add diagnostic benefit in the diagnosis of nonspinal osteomyelitis. In this study, composed of 18 adults with documented musculoskeletal infection, most cases of musculoskeletal infection were equally conspicuous on both pre- and post-contrast MRI sequences. However, they noted that inflammatory changes related to abscess and septic joint were more conspicuous on post-contrast sequences in eight patients (44.4%) [14].

A recent study by Averill et al. [15] found that there was no statistical difference in sensitivity and specificity between pre- and post-contrast sequences in children with suspected osteomyelitis. However, they observed that diagnostic confidence of osteomyelitis was higher with the addition of post-contrast imaging and that the ability to definitively exclude or diagnose osteomyelitis was improved in 46% of cases. The authors concluded that the routine administration of gadolinium is not necessary when pre-contrast imaging is normal. Our study, which was performed independently and without knowledge of Averill’s work, supports those findings. Our study had a broader scope, as we evaluated the impact of intravenous gadolinium in both bone and soft-tissue infections and the clinical impact of identifying additional abscesses, which was not addressed in their study. Similar to Averill et al.’s [15] study, we also observed that intravenous gadolinium improves conspicuity of abscesses in the setting of musculoskeletal infections. However, the surgical significance of this observation is problematic because many of the additional abscesses identified with post-contrast imaging did not require surgical debridement, particularly those related to cellulitis/myositis.

Our study has several limitations. One, it is a retrospective review of existing MRI studies. Statistical calculations were based on final clinical diagnosis as the reference standard; this is problematic in a retrospective study because final diagnosis was often based on a combination of blood work, clinical presentation and findings, initial imaging interpretation, and conclusions drawn by a heterogeneous group of physicians with varying clinical expertise over the years. Two, our criteria for intraosseous abscess on post-contrast images were any central nonenhancing marrow that showed corresponding increased signal on pre-contrast fluid-sensitive sequences; this could have resulted in overcalling abscesses, as these criteria can overlap with the diagnosis of secondary osteonecrosis related to osteomyelitis. Three, conclusions derived from our data on gadolinium in the setting of septic arthritis are limited by the small sample size.

Conclusion

The use of intravenous gadolinium in the imaging work-up of nonspinal pelvic- and extrapelvic musculoskeletal infections in children should not be routine in all patients but should be limited and based on scientifically proven accuracy criteria. Studies that are entirely normal based on non-contrast MRI sequences should not empirically undergo intravenous gadolinium-enhanced sequences. Children with abnormal non-contrast MRI findings such as soft-tissue edema, marrow edema, or abnormal joint effusions should also undergo contrast-enhanced sequences to potentially improve conspicuity and define extent of inflammation. Post-contrast sequences are helpful in potentially upgrading severity of inflammation and defining additional marrow and soft-tissue abscesses when present, although the smaller size of these abscesses might not justify surgical intervention.